QLoRA_RAG_test / data /eval_dataset.json
Dongjin1203's picture
Initial commit - RFPilot experiment
9630ae8
{
"metadata": {
"version": "1.0",
"description": "RFPilot 평가 데이터셋",
"created_by": "manual_annotation"
},
"in_distribution": [
{
"query": "사업 μ œμ•ˆμ„œ 제좜 λ§ˆκ°μΌμ€ μ–Έμ œμΈκ°€μš”?",
"expected_answer": "2024λ…„ 3μ›” 15μΌκΉŒμ§€μž…λ‹ˆλ‹€.",
"category": "deadline",
"expected_type": "document",
"source_doc": "RFP_2024_001.hwp",
"metadata": {
"difficulty": "easy"
}
},
{
"query": "μ œμ•ˆ μš”μ²­μ„œμ˜ 제좜 μ„œλ₯˜λŠ” λ¬΄μ—‡μΈκ°€μš”?",
"expected_answer": "κΈ°μˆ μ œμ•ˆμ„œ, κ°€κ²©μ œμ•ˆμ„œ, μ‚¬μ—…μžλ“±λ‘μ¦, νšŒμ‚¬μ†Œκ°œμ„œκ°€ ν•„μš”ν•©λ‹ˆλ‹€.",
"category": "requirements",
"expected_type": "document",
"source_doc": "RFP_2024_001.hwp",
"metadata": {
"difficulty": "medium"
}
},
{
"query": "사업 μ˜ˆμ‚° 규λͺ¨λŠ” μ–Όλ§ˆμΈκ°€μš”?",
"expected_answer": "총 5μ–΅μ›μž…λ‹ˆλ‹€.",
"category": "budget",
"expected_type": "document",
"source_doc": "RFP_2024_002.hwp",
"metadata": {
"difficulty": "easy"
}
}
],
"out_distribution": [
{
"query": "ν•œκ΅­μ˜ μˆ˜λ„λŠ” μ–΄λ””μΈκ°€μš”?",
"expected_answer": "μ„œμšΈμž…λ‹ˆλ‹€.",
"category": "general_knowledge",
"expected_type": "out_of_scope",
"metadata": {
"difficulty": "easy"
}
},
{
"query": "νŒŒμ΄μ¬μ—μ„œ λ¦¬μŠ€νŠΈμ™€ νŠœν”Œμ˜ μ°¨μ΄λŠ” λ¬΄μ—‡μΈκ°€μš”?",
"expected_answer": "λ¦¬μŠ€νŠΈλŠ” κ°€λ³€(mutable)이고, νŠœν”Œμ€ λΆˆλ³€(immutable)μž…λ‹ˆλ‹€.",
"category": "programming",
"expected_type": "out_of_scope",
"metadata": {
"difficulty": "medium"
}
}
]
}